This paper is addressed to the question of what rules the government should use when evaluating the efficiency gains from public sector projects in a tax-distorted economy. Our purpose will be to synthesize a variety of propositions which have appeared in the recent literature on "applied welfare economics". Of these, three might be singled out as being indicative of the various strands of thought. The first, due originally to Hotetling [9] but since developed by Harberger [6, 8] and Lesourne [10], states that when measuring welfare changes due to government policy changes (e.g. taxes), one must look at changes induced in the allocation of all goods and factors which are sold at distorted prices (prices above marginal costs). The second, also associated with Harberger [7] but found in Dasgupta, Sen and Marglin [4] and Sen [15], states that when the government undertakes a project which directly involves the use of goods or factors purchased or sold on distorted markets, it ought to account for the distortions by using a shadow price for the good or factor which is a weighted average of the demand and supply prices (the weights being determined by the response of supply and demand to the use of the market by the government). This is strictly a partial equilibrium approach since it does not require the government to take into account any distortions except those of the markets which the project directly uses. The third proposition is due to Diamond and Mirrlees [5]. It states that, under certain conditions, when the government sets all taxes optimally it should use as shadow prices in the public sector, private sector producer prices so as to achieve overall production efficiency in the economy. A simple general equilibrium model is employed here to illustrate the relationships amongst these propositions and the circumstances under which each is valid. A more general expression for cost-benefit analysis rules will be developed which will include the above rules as special cases.